Given other comments in this thread, it means (1) viewing the original work (in this case, the source code), (2) making superfluous changes to avoid literal copying (changing variable names, find-replace company name, etc), and (3) passing this off as original.
It sounds like it's not "You made a knockoff, that's 'non-literal copying,'", it's "You made a literal copy, changed a couple of things, and called it yours."
Structure, sequence, and organization of a computer program -- not just code -- is subject to copyright per Whelan v. Jaslow. Note that the Whelan standard became relevant again after Oracle v. Google.
I'm certainly not a lawyer, but I would definitely like to know what's the difference between "structure, sequence and organization of a computer program" and the algorithm behind that program? As I understand, algorithm is in the realm of ideas, which are not copyrightable.
It would be sad if courts accepted a technically ambiguous approach effectively deciding arbitrarily which case is what.
I thought this case was about trade secrets and NDA (and/or non compete).
I read this as the following example: So KFC has a secret blend of herbs and spices. If you worked there and stole a bag of the mix and sold it, it is pretty obviously theft. If you gave the recipe to someone else and they used it exactly, it is still theft of the secret as you can reproduce it exactly. Let's say instead that you use 5% less paprika in your mix but all other things are the same. This is a non literal copy but is still fully built on the trade secret.
It sounds like it's not "You made a knockoff, that's 'non-literal copying,'", it's "You made a literal copy, changed a couple of things, and called it yours."